public inbox for linux-um@lists.infradead.org
 help / color / mirror / Atom feed
From: Christoph Hellwig <hch@lst.de>
To: Jens Axboe <axboe@kernel.dk>,
	"Martin K. Petersen" <martin.petersen@oracle.com>
Cc: "Richard Weinberger" <richard@nod.at>,
	"Anton Ivanov" <anton.ivanov@cambridgegreys.com>,
	"Johannes Berg" <johannes@sipsolutions.net>,
	"Josef Bacik" <josef@toxicpanda.com>,
	"Ilya Dryomov" <idryomov@gmail.com>,
	"Dongsheng Yang" <dongsheng.yang@easystack.cn>,
	"Roger Pau Monné" <roger.pau@citrix.com>,
	linux-um@lists.infradead.org, linux-block@vger.kernel.org,
	nbd@other.debian.org, ceph-devel@vger.kernel.org,
	xen-devel@lists.xenproject.org, linux-scsi@vger.kernel.org
Subject: [PATCH 02/12] block: take io_opt and io_min into account for max_sectors
Date: Wed, 29 May 2024 07:04:04 +0200	[thread overview]
Message-ID: <20240529050507.1392041-3-hch@lst.de> (raw)
In-Reply-To: <20240529050507.1392041-1-hch@lst.de>

The soft max_sectors limit is normally capped by the hardware limits and
an arbitrary upper limit enforced by the kernel, but can be modified by
the user.  A few drivers want to increase this limit (nbd, rbd) or
adjust it up or down based on hardware capabilities (sd).

Change blk_validate_limits to default max_sectors to the optimal I/O
size, or upgrade it to the preferred minimal I/O size if that is
larger than the kernel default if no optimal I/O size is provided based
on the logic in the SD driver.

This keeps the existing kernel default for drivers that do not provide
an io_opt or very big io_min value, but picks a much more useful
default for those who provide these hints, and allows to remove the
hacks to set the user max_sectors limit in nbd, rbd and sd.

Note that rd picks a different value for the optimal I/O size vs the
user max_sectors value, so this is a bit of a behavior change that
could use careful review from people familiar with rbd.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-settings.c |  7 +++++++
 drivers/block/nbd.c  |  2 +-
 drivers/block/rbd.c  |  1 -
 drivers/scsi/sd.c    | 29 +++++------------------------
 4 files changed, 13 insertions(+), 26 deletions(-)

diff --git a/block/blk-settings.c b/block/blk-settings.c
index effeb9a639bb45..a49abdb3554834 100644
--- a/block/blk-settings.c
+++ b/block/blk-settings.c
@@ -153,6 +153,13 @@ static int blk_validate_limits(struct queue_limits *lim)
 		if (lim->max_user_sectors < PAGE_SIZE / SECTOR_SIZE)
 			return -EINVAL;
 		lim->max_sectors = min(max_hw_sectors, lim->max_user_sectors);
+	} else if (lim->io_opt) {
+		lim->max_sectors =
+			min(max_hw_sectors, lim->io_opt >> SECTOR_SHIFT);
+	} else if (lim->io_min &&
+		   lim->io_min > (BLK_DEF_MAX_SECTORS_CAP << SECTOR_SHIFT)) {
+		lim->max_sectors =
+			min(max_hw_sectors, lim->io_min >> SECTOR_SHIFT);
 	} else {
 		lim->max_sectors = min(max_hw_sectors, BLK_DEF_MAX_SECTORS_CAP);
 	}
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index 22a79a62cc4eab..ad887d614d5b3f 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -1808,7 +1808,7 @@ static struct nbd_device *nbd_dev_add(int index, unsigned int refs)
 {
 	struct queue_limits lim = {
 		.max_hw_sectors		= 65536,
-		.max_user_sectors	= 256,
+		.io_opt			= 256 << SECTOR_SHIFT,
 		.max_segments		= USHRT_MAX,
 		.max_segment_size	= UINT_MAX,
 	};
diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index 26ff5cd2bf0abc..05096172f334a3 100644
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -4954,7 +4954,6 @@ static int rbd_init_disk(struct rbd_device *rbd_dev)
 	    rbd_dev->layout.object_size * rbd_dev->layout.stripe_count;
 	struct queue_limits lim = {
 		.max_hw_sectors		= objset_bytes >> SECTOR_SHIFT,
-		.max_user_sectors	= objset_bytes >> SECTOR_SHIFT,
 		.io_min			= rbd_dev->opts->alloc_size,
 		.io_opt			= rbd_dev->opts->alloc_size,
 		.max_segments		= USHRT_MAX,
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index f6c822c9cbd2d3..3dff9150ce11e2 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -3593,7 +3593,7 @@ static int sd_revalidate_disk(struct gendisk *disk)
 	struct request_queue *q = sdkp->disk->queue;
 	sector_t old_capacity = sdkp->capacity;
 	unsigned char *buffer;
-	unsigned int dev_max, rw_max;
+	unsigned int dev_max;
 
 	SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp,
 				      "sd_revalidate_disk\n"));
@@ -3675,34 +3675,15 @@ static int sd_revalidate_disk(struct gendisk *disk)
 	else
 		blk_queue_io_min(sdkp->disk->queue, 0);
 
-	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
-		q->limits.io_opt = logical_to_bytes(sdp, sdkp->opt_xfer_blocks);
-		rw_max = logical_to_sectors(sdp, sdkp->opt_xfer_blocks);
-	} else {
-		q->limits.io_opt = 0;
-		rw_max = min_not_zero(logical_to_sectors(sdp, dev_max),
-				      (sector_t)BLK_DEF_MAX_SECTORS_CAP);
-	}
-
 	/*
 	 * Limit default to SCSI host optimal sector limit if set. There may be
 	 * an impact on performance for when the size of a request exceeds this
 	 * host limit.
 	 */
-	rw_max = min_not_zero(rw_max, sdp->host->opt_sectors);
-
-	/* Do not exceed controller limit */
-	rw_max = min(rw_max, queue_max_hw_sectors(q));
-
-	/*
-	 * Only update max_sectors if previously unset or if the current value
-	 * exceeds the capabilities of the hardware.
-	 */
-	if (sdkp->first_scan ||
-	    q->limits.max_sectors > q->limits.max_dev_sectors ||
-	    q->limits.max_sectors > q->limits.max_hw_sectors) {
-		q->limits.max_sectors = rw_max;
-		q->limits.max_user_sectors = rw_max;
+	q->limits.io_opt = sdp->host->opt_sectors << SECTOR_SHIFT;
+	if (sd_validate_opt_xfer_size(sdkp, dev_max)) {
+		q->limits.io_opt = min_not_zero(q->limits.io_opt,
+				logical_to_bytes(sdp, sdkp->opt_xfer_blocks));
 	}
 
 	sdkp->first_scan = 0;
-- 
2.43.0



  parent reply	other threads:[~2024-05-29  5:05 UTC|newest]

Thread overview: 43+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-05-29  5:04 convert the SCSI ULDs to the atomic queue limits API Christoph Hellwig
2024-05-29  5:04 ` [PATCH 01/12] ubd: untagle discard vs write zeroes not support handling Christoph Hellwig
2024-05-29  8:00   ` Damien Le Moal
2024-05-30 19:44   ` Bart Van Assche
2024-05-29  5:04 ` Christoph Hellwig [this message]
2024-05-29  8:05   ` [PATCH 02/12] block: take io_opt and io_min into account for max_sectors Damien Le Moal
2024-05-30 19:47   ` Bart Van Assche
2024-05-30 19:48   ` Ilya Dryomov
2024-05-31  5:54     ` Christoph Hellwig
2024-05-31  6:48       ` Ilya Dryomov
2024-05-31  6:56         ` Christoph Hellwig
2024-05-29  5:04 ` [PATCH 03/12] sd: simplify the ZBC case in provisioning_mode_store Christoph Hellwig
2024-05-29  8:07   ` Damien Le Moal
2024-05-30 19:48   ` Bart Van Assche
2024-05-29  5:04 ` [PATCH 04/12] sd: add a sd_disable_discard helper Christoph Hellwig
2024-05-29  8:10   ` Damien Le Moal
2024-05-30 19:50   ` Bart Van Assche
2024-05-29  5:04 ` [PATCH 05/12] sd: add a sd_disable_write_same helper Christoph Hellwig
2024-05-29  8:12   ` Damien Le Moal
2024-05-30 19:51   ` Bart Van Assche
2024-05-29  5:04 ` [PATCH 06/12] sd: simplify the disable case in sd_config_discard Christoph Hellwig
2024-05-29  8:13   ` Damien Le Moal
2024-05-30 20:02   ` Bart Van Assche
2024-05-29  5:04 ` [PATCH 07/12] sd: factor out a sd_discard_mode helper Christoph Hellwig
2024-05-29  8:14   ` Damien Le Moal
2024-05-29 21:11   ` Bart Van Assche
2024-05-29  5:04 ` [PATCH 08/12] sd: cleanup zoned queue limits initialization Christoph Hellwig
2024-05-29  8:18   ` Damien Le Moal
2024-05-30 20:07   ` Bart Van Assche
2024-05-29  5:04 ` [PATCH 09/12] sd: convert to the atomic queue limits API Christoph Hellwig
2024-05-29  8:23   ` Damien Le Moal
2024-05-30  9:16   ` John Garry
2024-05-31  5:48     ` Christoph Hellwig
2024-05-29  5:04 ` [PATCH 10/12] sr: " Christoph Hellwig
2024-05-29  8:25   ` Damien Le Moal
2024-05-29  5:04 ` [PATCH 11/12] block: remove unused " Christoph Hellwig
2024-05-29  8:28   ` Damien Le Moal
2024-05-30 20:08   ` Bart Van Assche
2024-05-31  9:14   ` John Garry
2024-05-29  5:04 ` [PATCH 12/12] block: add special APIs for run-time disabling of discard and friends Christoph Hellwig
2024-05-29  8:30   ` Damien Le Moal
2024-05-30 20:09   ` Bart Van Assche
2024-05-31  9:13   ` John Garry

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240529050507.1392041-3-hch@lst.de \
    --to=hch@lst.de \
    --cc=anton.ivanov@cambridgegreys.com \
    --cc=axboe@kernel.dk \
    --cc=ceph-devel@vger.kernel.org \
    --cc=dongsheng.yang@easystack.cn \
    --cc=idryomov@gmail.com \
    --cc=johannes@sipsolutions.net \
    --cc=josef@toxicpanda.com \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=linux-um@lists.infradead.org \
    --cc=martin.petersen@oracle.com \
    --cc=nbd@other.debian.org \
    --cc=richard@nod.at \
    --cc=roger.pau@citrix.com \
    --cc=xen-devel@lists.xenproject.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox